Subject Memory usage problems
Author William L. Thomson Jr.
I have a database that is consuming large portions of memory and
eventually eats up all the physical memory and swap in the machine.

I am running Fb 1.0.3 SS on RedHat Linux 9.

At some point I get the following errors in the log file

gds__alloc: memory pool corrupted

Which shows up again a 16 seconds later.

Then I get a second after the last error

gds__free: pool corrupted
gds__alloc: memory pool corrupted

Which eventually starts to deny connections and I get the following

INET/inet_error: connect errno = 111

The ibserver process does not bring down the machine. However it does
get itself to a point where I must restart Firebird in order to use the
database again. Once ibserver is stopped, I get my memory back and I can
bring the service up again without rebooting the machine.

What really shocks me is the database is not even 2Mb in size yet. Yet I
have seen it consume 256Mb of physical memory and another 256Mb of swap.
I have the catch page set to 4096 and forced writes is turned on.

I came across a mention of this in bug # 114840

Supposedly it was a bug that has been fixed. I am not sure why I am
running into it.

I am aware of a few transactions that commit or rollback is not being
called on. Basically making them open transactions, however I believe
they are not using much memory. I am working on correct that. However I
do not believe that will solve my memory problems log in by Firebird.

Anyone care to comment on this or provide suggestions as to how I can
get Firebird to release and manage it's memory properly? I need to get
this problem addressed ASAP.

Thank you

William L. Thomson Jr.
Support Group
Obsidian-Studios, Inc.